CHUNG-HYE HAN and MARIBEL ROMERO
THE SYNTAX OFWHETHER/Q . . . ORQUESTIONS: ELLIPSIS
COMBINED WITH MOVEMENT
ABSTRACT. In this paper, we argue that the syntax ofwhether/Q . . . orquestions involves both movement ofwhether/Qand ellipsis of the type that has been argued to exist foreither . . . orconstructions. Three arguments are presented: (i) Englishwhether/Q . . . orquestions present at the same time movement characteristics (sensitivity to islands) and ellipsis traits (focus pattern on the disjuncts); (ii) crosslinguistic data on the surface string syn- tax of Subject-Object-Verb (SOV) languages support the ellipsis plus movement account in general and, thus, indirectly also for English; and (iii) certain asymmetries between whether/Q...orandeither...orare resolved, permitting a unified account of the two types of constructions.
1. INTRODUCTION
In English, a question may have a disjunctive phrase explicitly providing the choices that the question ranges over. For example, in (1), the disjunc- tionor notindicates that the choice is between the positive and the negative polarity for the relevant proposition, as spelled out in the yes/no (yn)- question reading (2) and in the answers (2a,b). Another example is (3).
The disjunction in (3) can be understood as providing the choices that the question ranges over, hence giving rise to the alternative (alt-)reading in (4) and eliciting the answers in (4a,b) (cf. Karttunen 1977 and Higginbotham 1993 for the semantics ofyn/alt-questions).
(1) Did John eat beans or not?
(2) yn-reading: “Is it true or false that John ate beans?”
a. Yes, John ate beans.
We thank Rajesh Bhatt for extensive discussion on several aspects of this paper. This paper has also benefitted from discussions and comments at various stages from Lorna Fadden, Nancy Hedberg, Tony Kroch, Dean Mellow, Bill Poser, Beatrice Santorini, and the audiences at NELS 32 and the Dept. of Linguistics colloquium at Simon Fraser University in Fall 2001. We also wish to thank the anonymous reviewers for their critical comments that helped us reshape and improve this paper. All remaining errors are ours. Finally, Han acknowledges SSHRC Standard Research Grant #410-2003-0544 for support when this paper was written.
Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-139190
b. No, John didn’t eat beans.
(3) Did John eat beans or rice?
(4) alt-reading: “Which of these two things did John eat: beans or rice?”
a. John ate beans.
b. John ate rice.
Questions with an associated disjunction phrase can also occur, of course, in embedded contexts. (5) and (6) are the embedded counterparts of (1) and (3) respectively, with the indicated relevant readings:
(5) I wonder [whether John ate beans or not].
a. yn-reading: “I wonder whether it is true or false that John ate beans.”
(6) I wonder [whether John ate beans or rice].
a. alt-reading: “I wonder which of these two things John ate: beans or rice.”
The syntax of questions with an overt associated disjunction phrase is the topic of this paper. Note that (3) and (6) also have a yn-reading, spelled out in (7). Under this reading, the question choice does not range over the overt disjuncts beansand rice, but over a positive and negative polarity not explicitly expressed. The derivation of thisyn-question reading without an overt associated disjunction is beyond the scope of this paper;
however, in section 6.2, we will briefly sketch possible analyses of it that are compatible with the main claims of this paper.
(7) yn-reading for (3) and (6):
“Is it true or false that John ate any of these two things, beans or rice?”
a. Yes, John ate beans or rice.
b. No, John didn’t eat beans or rice.
Throughout this paper, we will call both matrix and embedded ques- tions with an associated disjunctive phrase ‘whether/Q . . . or construc- tions’. Q corresponds to covert whether. Positing a covert whether in matrix questions is motivated by the fact that there are languages that allow
overtwhetherin matrixyn- or alt-questions. Examples from Early Modern English and Yiddish are given in (8) and (9).
(8) Early Modern English (examples taken from Radford (1997, p.
295)
a. Whetherhad you rather lead mine eyes or eye your master’s heels?
‘Would you rather lead my eyes or eye your master’s heels?’
(Mrs. Page,Merry Wives of Windsor, III.ii)
b. Whetherdost thou profess thyself a knave or a fool?
‘Do you profess yourself (to be) a knave (= scoundrel) or a fool?’ (Lafeu,All’s Well that Ends Well, IV.v)
(9) Yiddish a. (Tsi)
whether
reykhert smokes
Miryam?
Miryam
‘Does Miryam smoke?’
b. Shmuel Shmuel
hot has
mikh me
gefregt, asked
tsi whether
Miryam Miryam
reykhert.
smokes
‘Shmuel asked me whether Miryam smokes’.
There have been two main approaches to the syntax of disjunctiveOp- erator . . . orconstructions in the literature. Larson (1985) assimilates the syntax ofwhether/Q . . . orconstructions toeither . . . orconstructions as in (10), arguing that in both,eitherorwhether/Qis base-generated adjacent to the disjunctive phrase and that it undergoes movement.
(10) Either John ate beans or rice.
Schwarz (1999), on the other hand, showed that the syntax ofeither . . . or can be accounted for better if we assume that the second disjunct is a clause with a particular type of ellipsis, namely gapping. Unfortunately, he was not able to extend the ellipsis analysis towhether/Q . . . or, because he found certain asymmetries betweenwhether/Q . . . orstructures and the type of ellipsis allowed ineither . . . or.
In this paper, we propose that the syntax of whether/Q . . . or ques- tions involves ellipsis of the type that has been argued to exist foreither . . . orconstructions, as illustrated in (11). We also argue, with Larson, that whether/Qundergoes movement. The representation of alt-questions that
we will defend in this paper is illustrated in (12). We attribute the apparent asymmetries between whether/Q . . . orand either . . . or constructions to the fact thatwhether/Qis awh-phrase that can undergo movement, while eitheris not.
(11)a. Either John ate beans or rice.
b. Either [John ate beans] or [John ate rice]
(12)a. (Q/whether) did John eat beans or rice?
b. (Q/whether)i didti [John eat beans] or [John eat rice]
Three arguments for our combined ellipsis/movement analysis of whether/Q . . . orquestions will be presented: (i) Englishwhether/Q . . . or questions present at the same time movement characteristics (sensitivity to islands) and traits that follow naturally under an ellipsis account (fo- cus pattern on the disjuncts); (ii) crosslinguistic data on the surface string syntax of Subject-Object-Verb (SOV) languages show that alt-readings in- volve ellipsis and movement (or a parallel scoping mechanism) in these languages, hence indirectly supporting the same analysis for English; (iii) finally, Schwarz’s asymmetries betweenwhether/Q . . . orandeither . . . or in English are resolved, permitting a unified account of the two types of constructions.
This paper is organized as follows. In sections 2 and 3, we will briefly present the main points of Larson’s movement account and Schwarz’s el- lipsis account. We will then further argue that whether/Qis awh-phrase undergoing movement to [Spec, CP], in section 4. This movement is sub- ject to island constraints, just like any other wh-movement. In section 5, we argue that in conjunction with whether/Q-movement, whether/Q . . . or questions involve ellipsis of the type that has been argued to exist for either. . . ordeclaratives. Just as ineither. . . or declaratives, ellipsis in whether/Q . . . orquestions is constrained to be restricted to a finite clause.
Our main argument for ellipsis in whether/Q . . . or questions will come from languages with canonical SOV word order – Hindi and Korean. We will also argue that the focus pattern found in English alt-questions can only be given a natural explanation under the ellipsis account. In section 6, we show how the combination ofwhether/Q-movement and ellipsis derives English alt-questions, thereby giving a unified account ofeither . . . orand whether/Q . . . orconstructions as well as accounting for their asymmetries.
We conclude with some discussion onwhether/Q . . . orquestions involving right-node raising, in section 7.
2. LARSON’SMOVEMENTACCOUNT
According to Larson (1985), in sentences witheither . . . or,eitherorigin- ates adjacent to a disjunctive phrase, and moves to higher up in the clause, determining the scope of disjunction. For instance, in (13a), eitherorigin- ates adjacent torice or beans, moving to the surface position adjacent to VP. In (13b), either originates adjacent to rice or beans, moving to the surface position adjacent to IP.
(13)a. John either ate rice or beans.
John eitheri ateti [NPrice] or [NPbeans]
b. Either John ate rice or beans.
Eitheri John ateti [NPrice] or [NPbeans]
Larson extends the movement analysis of either . . . or constructions to whether/Q . . . or questions. He argues that a question with associated disjunction has a question operator: whetheror nullQ. This operator ori- ginates from a disjunction phrase and moves to [Spec, CP], marking the scope of disjunction. Moreover, this type of question may have an overt or an unpronounced disjunction phrase or not. If the disjunction phrase from which the whether/Qoriginates is the overt or the unpronouncedor not, then the yn-question reading is derived. Otherwise, the alt-question reading is derived. For instance, the question in (14) can have either ayn- question reading or an alt-question reading. Under theyn-question reading, the whether/Qoperator originates from or not and moves to [Spec, CP], and the disjunction phrase coffee or tea is associated with another oper- ator, as represented in (14a). Under the alt-question reading, thewhether/Q operator originates from the disjunction phrasecoffee or teaand moves to [Spec, CP], as represented in (14b).
(14) Did John drink coffee or tea?
a. yn-question:
Opi (ti or not) [did John drink [Opj coffee or tea]]
{John drank coffee or tea, John didn’t drink coffee or tea}
b. alt-question:
Opi [did John drink [ti coffee or tea]]
{John drank coffee, John drank tea}
As noted in Larson, supporting evidence for the proposal that whether/Q moves from a disjunction phrase to [Spec, CP] comes from
the fact that questions that have the intended associated disjunction phrase inside an island do not have the alt-question reading available.
(15) Do you believe the claim that Bill resigned or retired?
a. yn-question:
Opi (ti or not) [do you believe [NPthe claim that Bill resigned or retired]]
b. ∗alt-question:
Opi [do you believe [NP the claim that Bill [ti resigned or retired]]]
In (15), the disjunctive phraseresigned or retiredis inside a complex NP.
The alt-question reading is not available since the empty operator would have to move out of an island to generate this reading, as represented in (15b). But the yn-question reading is available, since under this reading the empty operator is moving from the unpronouncedor not, which is not inside an island. This reading is represented in (15a).
3. SCHWARZ’S ELLIPSIS ACCOUNT
Schwarz (1999) argues that the syntax ofeither . . . orconstructions can be assimilated to the syntax of coordinate constructions that involvegapping, a type of ellipsis. Gapping originally refers to the grammatical process which is responsible for the deletion of a verb in the second coordinate of a conjunctive coordination under identity with the first coordinate, as in (16a) (Ross 1970). The deleted material in the second coordinate is called gap, and the materials in the second coordinate that have not been deleted are called remnants. Schwarz points out that gaps may contain more than just a verb, as shown in (16b), and argues that this fact is comparable with the idea thateither . . . orconstructions involve gapping.
(16)a. Tom has a pistol and Dick a sword.
[Tom has a pistol] and [Dick has a sword]. (Schwarz 1999, ex.
30a)
b. Bill devoured the peaches quickly and Harry slowly.
[Bill devoured the peaches quickly] and
[Harry devoured the peaches slowly]. (Schwarz 1999, ex. 33a) According to Schwarz, ineither . . . orconstructions, eithermarks the left periphery of the first disjunct. This means that disjunctive declaratives with
either occurring at the edge of a VP or an IP have VP or IP disjunctive structures, and so some materials in the second disjunct can be deleted under identity with the first disjunct. This is illustrated in (17).1
(17)a. John either ate rice or beans.
John either [VP ate rice] or [VP ate beans] (Schwarz 1999, ex.
28a)
b. Either John ate rice or beans.
Either [IP John ate rice] or [IP John ate beans] (Schwarz 1999, ex. 28b)
A compelling piece of supporting evidence for gapping analysis of either . . . orconstructions comes from unbalanced disjunction. It is gener- ally accepted in the literature that the conjuncts in the Coordinate structure to which gapping applies must be parallel to each other (Hankamer 1971;
Stillings 1975; Hudson 1976). If the parallelism constraint is violated, then the second conjunct would end up with a dangling remnant which lacks a correlate in the first conjunct. Schwarz notes that dangling remnants are prohibited in gapping constructions, aswith meand clumsilyin (18), and shows that they are prohibited ineither . . . orconstructions as well, asoff andupin (19).
(18)a. ∗Some talked about politics and others with me about music.
[some talked about politics] and [others talked with me about music]. (Schwarz 1999, ex. 40b)
b. ∗John dropped the coffee and Mary clumsily the tea.
[John dropped the coffee] and [Mary clumsily dropped the tea].
(Schwarz 1999, ex. 41b)
(19)a.??Either this pissed Bill or Sue off.
Either [this pissed Bill] or [this pissed Sue off]. (Schwarz 1999, ex. 43a)
1 Given that gapping usually leaves more than one remnant, it might be surprising that Schwarz considers examples in (17) to be gapping constructions. Schwarz points out that there is no known argument to the effect that gapping must leave more than one remnant, and presents split coordinations, as in (i), as instances of gapping with one remnant, citing Hudson (1976), Neijt (1979), Reinhart (1991), and Johnson (1996).
(i)a. John bought a book yesterday, and a newspaper. (Schwarz 1999, ex. 36a) b. John came, and Bill (too). (Schwarz 1999, ex. 36b)
b.??Either they locked you or me up.
either [they locked you] or [they locked me up]. Schwarz 1999, ex. 43c)
Larson’s movement account on the other hand has nothing to say about the fact that examples with dangling remnants are degraded.
Schwarz however did not extend the gapping analysis to alt-questions because gapping allowed in whether/Q . . . orconstructions appears to be different from either . . . or constructions. First of all, whether/Q . . . or constructions seem to allow dangling remnants, unlike either . . . or con- structions and other coordinate constructions with gapping.
(20)a. Did this piss Bill or Sue off?
b. Did she turn the test or the homework in?
The questions in (20) can both have the alt-question reading. But if we were to apply the gapping analysis to these questions, then we would end up with dangling remnants, which were prohibited from other gapping constructions.
Second, whether/Q . . . or constructions behave differently from other gapping constructions in that while elided materials cannot spread across matrix and embedded finite clauses, as shown in (21), they seem to be able to inwhether/Q . . . orconstructions, as in (22).
(21)a. ∗The first letter says that you should pay tax and the second letter V.A.T.
[The first letter says that you should pay tax] and [the second letter says [that you should pay V.A.T]]. (Schwarz 1999, ex.
61a)
b.??Either Bill said that Mary was drinking or playing video games.
Either [Bill said that Mary was drinking] or [Bill said [that Mary was playing video games].
(22)a. Did Bill say that Mary was drinking or playing video games?
b. Did John say that Bill retired or resigned?
c. Did John claim that Bill drank coffee or tea?
The questions in (22) all have the alt-question reading available. If this reading was derived via gapping in the second disjuncts in (22), then the el- lipsis would spread beyond a finite clause boundary. But this is impossible in other gapping constructions.
In sum, Larson defends a pure movement account ofOp . . . orconstruc- tions and Schwarz proposes a pure ellipsis analysis of either . . . or. We argue that both ellipsis and movement are needed to generate whether/Q . . . orquestions. We claim thatwhether/Q . . . orquestions have a structure parallel to either . . . or constructions, except for the additional fact that whether/Qmoves to [Spec, CP]. Our arguments for this combined move- ment/ellipsis approach are presented in the next two sections. In section 4, we argue that whether is a wh-phrase and that, as such, it undergoes wh-movement. In section 5, we show that ellipsis is needed to explain the syntactic properties of alt-questions in SOV languages like Hindi and Korean, and that the focus intonation in English alt-questions follows naturally from an ellipsis account.
4. Whether/Q IS Awh-PHRASE
One compelling piece of evidence that supports the idea thatwhetheris in [Spec, CP] comes from the fact that there are languages that allowwhether to cooccur with a complementizer. For example, this is attested in Middle English. The examples in (23) are from Penn-Helsinki Parsed Corpus of Middle English (Kroch and Taylor 2000).
(23)a. for men weten nou+gtwhe+ter +tathe leue+t or is dede.
‘For people don’t know whether he is alive or is dead’
(cmbrut3,90.2727)
b. andwhe+ter +tathe shal ascape or dye, at Godes wil most hit be.
‘and whether he escapes or dies, it must be according to God’s will’ (cmbrut3,107.3225)
c. aske hymwhe+ter +tater+te ys herre +ten Heuen?
‘ask him whether Earth is higher than Heaven?’
(cmmirk,10.273)
d. for Seynt Ierome sei+twhe+tur +tathe ete, drynke, or slepe, or what-euer els +tat he dothe,
‘for Saint Jerome says that, whether he eats, drinks, or sleep;s, or whatever else that he does’ (cmroyal,18.162)
e. it weere for to witen whether thatalle thise thinges make or conjoynene as a maner body of blisfulnesse by diversite of parties or membres, . . .
‘it would be to know whether all these things make or con- join as a manner a body of blissfulness by diversity of parts or members . . . ’ (cmboeth,433.C1.189)
It turns out that Middle English also allows awh-phrase to cooccur with a complementizer, as in (24).
(24) First the behoueth to knowewhy thatsuche a solitary lyf was ordeyned.
‘First, it behooves thee to know why such a solitary life was ordained.’ (cmaelr4-s0.m4,13)
b. . . . he wiste wel hymselfwhat thathe wolde answere . . .
‘he himself knew well what he would answer’ (cmctmeli- s0.m3,75)
c. Now shall ye understondein what manere thatsynne wexeth or encreesseth in man.
‘Now you shall understand in what manner sin grows or increases in man.’ (cmctparss0.m3,390)
d. I pray you telle mewhat knyght thatye be
‘Please tell me what knight you are’ (cmmalory-20.m4,4655) The same facts obtain in Belfast English (Alison Henry, p.c.).
(25)a. ?John wonderswhether thatMary ate beans.
b. John wonderswhich sandwich thatMary ate.
This suggests that whether should not be classified as a complementizer, but as awh-phrase, occupying [Spec, CP].
Furthermore, Larson (1985) points out, citing Jespersen (1909–1949), that whetherdeveloped historically as thewh-counterpart ofeither, and it originally meant ‘which of either A or B’. If whether is a wh-phrase, it is reasonable to assume that it and its covert version Qend up in [Spec, CP] by movement, just like any otherwh-phrases. Moreover, just like any otherwh-movement,whether/Q-movement is not clause-bound, but rather island-bound. The contrast in grammaticality between (26) and (27) illus- trates this fact. The movement of Qout of a complement clause is fine, hence the alt-reading in (26) is possible. But a complex NP is an island and so the movement ofQin (27) is impossible, ruling out the alt-reading.
(26)a. Did John say that Bill resigned or retired? (alt-reading)
b. Qi Did John say [that Billti [resigned or retired]]?
(27)a. ∗Did John believe the claim that Bill resigned or retired? (alt- reading)
b. ∗QiDid John believe [the claim that Billti[resigned or retired]]?
5. ELLIPSIS INADDITION TOMOVEMENT INALT-QUESTIONS
We have argued, in line with Larson (1985), that whether/Qmoves from the left edge of the disjunction phrase to [Spec, CP], likewh-phrases move.
Once we allow whether/Q-movement, it should be possible for an alt- question as in (28a) to have the parsing in (28b), where Q is extracted from an NP disjunctive phrase, taking us back to the Larsonian derivation ofOp . . . orconstructions. After all,eithercan occur at the left edge of an NP, marking an NP disjunction, as in (29). Hence, the source position for whether/Qin (28a) could very well be the positioneitheroccupies in (29).
(28)a. Did John drink coffee or tea?
b. Qi Did John drinkti [coffee or tea]
(29) John drank either [coffee or tea].
We have seen however in section 3 that a possible derivation for declar- atives with or involves clausal disjunction, either at IP or VP level, with ellipsis in the second clause as in (30).2 Given the cross-categorial status of or, if or can disjoin clauses in declaratives, there is no reason why it should not do so in questions. What this means is that in principle, for an alt-question as in (28a), the grammar would not only generate (28b), but also (31) with both movement ofQand ellipsis in the second disjunctive clause.
(30) Either [John drank coffee or he drank tea].
(31) Qi Didti [John drink coffee or John drink tea]?
The question then is whether one derivation wins over the other, and if so, which one. In the next two subsections, we will provide arguments
2 Under the VP-internal subject hypothesis, VP is clausal.
that support the derivation of the type in (31) – clausal disjunction plus ellipsis – for alt-questions. In subsection 5.1, we present evidence from cross-linguistic data that strongly supports our proposal that whether/Q . . . orconstructions involve disjunction at clausal level, IP or VP, with con- sequent ellipsis in the second clause. In subsection 5.2, we show that the focus pattern in alt-questions found in English can naturally be explained only under the ellipsis account.
5.1. Cross-Linguistic Support
Word order facts in languages that have canonical SOV word order provide evidence that the syntax of alt-questions involves clausal disjunction with ellipsis. We will examine two SOV languages here: Hindi and Korean.
Moreover, scope marking data from Hindi supports the thesis that, besides ellipsis, a scoping mechanism is at work in alt-readings, just as in wh- questions. Finally, case marking in Korean further supports the ellipsis view.
5.1.1. Hindi
Hindi provides two arguments for the combined ellipsis/movement ap- proach advocated in this paper. The first argument concerns the ellipsis part of the analysis; the second argument reflects the need of an extra scoping mechanism beyond the size of the coordinated constituents (cf.
wh-movement in English).
Word order in Hindiyn-and alt-questions furnishes the first argument.
Hindi being an SOV language, the surface location of the finite verb gener- ally marks the right edge of the clause (and of the VP). This circumstance raises a prediction: if alt-readings are construed by disjoining two IPs (or possibly two VPs) and partially eliding the second IP (or VP) and if no further movement occurs, then the second disjunct should appear after the right boundary of the first IP (or VP). This prediction is borne out. Theyn- reading of a question such asDid Chandra drink coffee or tea?is expressed by having the NP disjunct coffee or teain the object position, as in (32).
This question can never have an alt-reading.
(32) (Kyaa) what
Chandra-ne Chandra-Erg
coffee coffee
yaa or
chai tea
pii?
drink-Pfv
‘Is it the case that Chandra drank coffee or tea?’ (yn-question) For the alt-reading to obtain,coffeeandteamust be separated by the verb drink, as in (33a):coffee is the object of the first disjunctive clause, and teaoccurs alone in the second disjunct. As predicted by our analysis, this
suggests that, in the alt-reading, yaa (‘or’) coordinates two full clausal constituents, rather than two NPs, in the second one of which the subject and the verb have been elided leaving the object chai (‘tea’) as the only remnant. This is represented in (33b).3
(33)a. (Kyaa) what
Chandra-ne Chandra-Erg
coffee coffee
pii drink-Pfv
yaa or
chai?
tea
‘Which of these two things did Chandra drink: coffee or tea?’
(alt question) a. (Kyaa)
what Q Q
[Chandra-ne Chandra-Erg
coffee coffee
pii]
drink-Pfv yaa or
[Chandra ne Chandra-Erg chai
tea pii]?
drink-Pfv
Let us now turn to the second argument. Based on the distribution of the wh-element kyaa, we will argue that alt-questions involve a scoping mechanism for whether/Q. Interestingly, this scoping mechanism is the same as the one used to give wider scope to regular wh-phrases in wh- questions.
To make the argument, note, first, that Hindi is a wh-in-situ language with LFwh-movement (overtly movedwh-phrases are generally assumed to be scrambled), as exemplified in (34) (Dayal 1996).
(34) Tum-ne you-Erg
us-ko he-Dat
kyaa what
diyaa?
give-Pfv
‘What did you give him?’
It has also been noted that LFwh-movement is not possible out of finite clauses in Hindi, as in (35).
(35) Tum you
jaante know
ho be
[ki that
us-ne he-Erg
kyaa what
kiyaa]
do-Pfv
‘You know what he did.’
∗‘What do you know he did?’
3 In principle, ayn-question reading should be available to (33a), as well as an alt- reading. But only alt-reading seems to be readily available to native speakers. This may be a pragmatic effect caused by disjoining full clauses in awhether/Qquestion. The same effect is attested in Englishwhetherquestions with two disjoined clauses: e.g.,I wonder whether John drank coffee or he drank tea.We will see later (cf. (37), (43b)) that ayn- question reading is available towhether/Qquestions when disjoined clauses are restricted to embedded contexts.
The way to obtain a direct question when the wh-phrase originates in an embedded finite clause is by using the wh-element kyaa ‘what’ in the matrix clause, where it acts as the scoping mechanism for the embedded wh-phrase, as in (36).4Note that the presence ofkyaais obligatory for the embeddedwh-phrase to attain matrix scope in this example. When there is no such embedding (e.g., in simple alt-questions like (33a) and in simple yn-questions like (32)),kyaais merely optional.5
(36) Jaun John
kyaa what
soctaa think
hai be-Pres
[ki that
merii Mary
kis-se who-Ins
baat talk
karegii]?
do-Fut
‘What does John think, who will Mary talk to?’
i.e., ‘Who does John think Mary will talk to?’
The issue we want to examine here is what happens with Hindi alt- questions with disjunction originating from an embedded clause, as in (37).
Do we coordinate two matrix clauses allowing for ellipsis to spread across a finite clause boundary, as in (38)? Or, in a way comparable to our hybrid proposal for English, doesorcoordinate only the two embedded IPs, as in (39), and an extra scoping mechanism is responsible for the matrix scope interpretation of the embeddedwhether/Q?
(37) Jaun John
kyaa what
sochtaa think
hai be-Pres
[ki that
Chandra-ne Chandra-Ern
coffee coffee
pii drink-Pfv thii
Past yaa or
chai]?
tea
English counterpart: ‘Does John think that Chandra drank coffee or tea?’ (alt-question)6
4 According to Dayal (1996, 2000), in examples such as (36), bothkyaain the matrix clause and kis-sein the embedded clause are regular wh-phrases, each undergoing LF movement to the [Spec, CP] of their respective clauses. Semantically,kyaais treated as quantifying over propositions and the embedded clause forms the restriction of this quanti- fication. This semantic mechanism ends up returning as possible answers the ‘propositions that John stands in thethinkrelation to and which furthermore are members of the indirect question’ (Dayal 1996, p. 7). In effect, the set of possible answers are: {John thinks that Mary will talk to Sue, John thinks that Mary will talk to Sita, John thinks that Mary will talk to Chandra . . . }. Thus, the embeddedwh-phrase obtains matrix scope interpretation, without actually having matrix scope structurally.
5 Note that while Hindi clauses in general have SOV word order, CP complement clauses occur to the right of the verb.
6 (37) can have either an alt-question reading or ayn-question reading. Here we are only concerned with the alt-reading. The alt-reading can be semantically derived equally from the structure in (38) and from the structure in (39).
(38) Juan John
kyaa what
Q Q
[sochtaa think
hai be-Pres
[CP ki that
Chandra-ne Chandra-Erg
coffee coffee pii
drink-Pfv thii]]
Past yaa or
[sochtaa hai think be Pres
[CP ki that
Chandra-ne Chandra-Erg
chai tea pii thii]]?
drink-Pfv Past
‘Which one is true: John thinks that Chandra drank coffee or John thinks that Chandra drank tea?’ (same reading as in (39)) (39) Jaun
John kyaa what
sochtaa think
hai be-Pres
[CP Q Q
ki that
[Chandra-ne Chandra-Erg
coffee coffee pii
drink-Pfv thii]
Past yaa or
[Chandra-ne Chandra Erg
chai tea
pii thii]]?
drink Pfv Past
‘Which does John think: that Chandra drank coffee or that Chandra drank tea?’ (same reading as in (38))
We argue that the correct structure for the alt-reading of (37) is (39), a hybrid ellipsis/scoping structure, for the reason thatkyaais obligatory for the alt-reading of (37) to obtain and that removingkyaa, as in (40), renders the alt-reading impossible. That is, the matrixyn-reading obtains both in (37) and in (40), no matter whetherkyaaappears or not, but the alt-reading can only obtain whenkyaais present.
(40) Jaun John
sochtaa think
hai be-Pres
[ki that
Chsnfts-ne Chandra-Erg
coffee coffee
pii drink-Pfv
thii Past yaa
or
chai]?
tea
∗Which does John think: that Chandra drank coffee or that he drank tea?’ (alt-question)
‘Is it the case that John thinks this: that Chandra drank coffee or tea?’ (yn-question)
If we were to assume a pure ellipsis account and take the structure in (38), the scope of disjunction would be in the matrix clause already, and, hence, the particlekyaawould be expected to be optional, as it is in (33a) and in the matrixyn-question reading of (40). Under the hybrid ellipsis/scoping account with the structure in (39), the need for kyaahere is explained as the need forkyaawith embeddedwh-phrases. Wh-phrases by themselves cannot take scope beyond their clause, and neither can gapping spread
across a finite clause, as the examples (41)–(42) with declaratives show.
Both with regular wh-phrases and with disjunctive wh-phrases, then, the presence of the extra scoping mechanism is necessary in order to give the embeddedwh-element a matrix scope interpretation.
(41) Sue Sue
aaluu potatoes
pasand like
kartii do-Hab
hai be-Pres
aur and
Martha Martha
pyaaz.
onions
‘Sue likes potatoes and Martha onions.’
(42) Sue Sue
sochtii think-Hab
hai be-Pres
ki that
mE I
Bill-ko Bill-Dat
pasand like
kartaa do-Hab
huN be-1Sg aur
and
Martha Martha
Tom-ko Tom-Dat
‘Sue thinks that I like Bill and Martha (likes) Tom.’
∗‘Sue thinks that I like Bill and Martha (thinks that I like) Tom.’
In fact, the scoping mechanism for embeddedwh-phrases and the scop- ing mechanism for embedded whether/Q in Hindi not only share the necessary presence of kyaa, but they are also equally subject to island constraints. For instance, (43a) is grammatical under theyn-reading, but it is impossible under the matrixwh-question reading. Similarly,kyaacannot mark the scope of whether/Q associated with a disjunction in an island.
That is why (43b) cannot be interpreted as an alt-question. It can only be interpreted as ayn-question.
(43)a. Ram-ko Ram-Dat
kyaa what
yeh this
baat thing
pataa known
hai be-Pres
ki that
Chandra-ne Chandra-Erg kyaa
what pii drink-Pfv
thii?
Past
‘Does Ram know what Chandra drank?’
∗‘What does Ram know the fact that Chandra drank?’
b. Ram-ko Ram-Dat
kyaa what
yeh this
baat thing
pataa known
hai be-Pres
ki that
Madhu-ne Madhu-Erg
chai tea pii
drink-Pfv thii Past
yaa or
coffee?
coffee
‘Does Ram know the fact that Madhu drank tea or Madhu drank coffee?’ (yn-question)
∗‘Which fact does Ram know: that Madhu drank tea or that Madhu drank coffee?’ (alt-question)
In summary, Hindi alt-questions are explained under the combined ellipsis/scoping account. First, word order facts suggest that Hindi alt- questions involve ellipsis. Second, given that this type of ellipsis (gapping) cannot spread across a finite clause boundary, a scoping mechanism with obligatorykyaais needed in order to give matrix scope interpretation to a disjunction originating in an embedded clause. This ellipsis/scoping ana- lysis for Hindi is parallel to the ellipsis/movement analysis we want to propose for English, hence lending crosslinguistic support to it.
5.1.2. Korean
Word order and case-marking facts support the ellipsis analysis of alt- questions in Korean. As in Hindi, Korean questions with an associated disjunctive phrase such as coffee or tea in the object position can only have ayn-question reading, as in (44).
(44) Chelswu-ka Chelswu-Nom
khophi-na coffee-or
cha-lul tea-Acc
masi-ess-ni?
drink-Past-Int
‘Is it the case that Chelswu drank coffee or tea?’ (yn-question) In order to obtain an alt-question reading, first of all, a different disjunctive connective,animyenwhich literally meansif not, must be used. In contrast to-naas in (44), animyencan only disjoin clause level constituents, and only has the meaning of exclusivity. Second, coffee must be part of the first disjunctive clause, and tea must be part of the second disjunctive clause. There are two ways to achieve this: (i) by disjoining two clauses (or possibly two VPs) without deleting the verb from either of the clauses (or VPs), as in (45a); or (ii) by disjoining two clauses (or two VPs) with deletion of the verb from the first clause (or first VP), as in (45b).
(45)a. Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
masi-ess-ni drink-Past-Int
animyen if-not
cha-lul tea-Acc masi-ess-ni?
drink-Past-Int
‘Which of these two things did Chelswu drink: coffee or tea?’
(alt-question) b. Chelswu-ka
Chelswu-Nom
khophi-lul coffee-Acc
animyen if-not
cha-lul tea-Acc
masi-ess-ni?
drink-Past-Int
‘Which of these two things did Chelswu drink: coffee or tea?’
(alt-question)
The syntax of (45a) as a disjunction of two clauses is straightforward.
Korean, however, is different from Hindi and English in that deletion tar- gets the verb in the first clause and not the one in the second clause, as the contrast between (45b)–(46) shows. This deletion pattern is the same that we encounter in gapping in Korean declaratives, as in (47a) and (47b):
(46) ∗Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
masi-ess-ni drink-Past-Int
animyen in-not
cha-lul?
tea-Acc
‘Which of these two things did Chelswu drink: coffee or tea?’
(alt-question) (47)a. Tol-i
Tol-Nom
Seoul-ey Seoul-in
kuliko and
Anoop-i Anoop-Nom
India-ey India-in
sal-ass-ta.
live-Past-Decl
‘Tol lived in Seoul and Anoo in India.’
b. ∗Toi-i Tol-Nom
Seoul-ey Seoul-in
sal-ass-ko live-Past-Conj
kuliko and
Anoop-i Anoop-Nom
India-ey.
India-in
‘Tol lived in Seoul and Anoop in India.’
One may think that the surface string of (45b) illustrates ellipsis from an unbalanced disjunction, as represented in (48).
(48) [Chelswu-ka Chelswu-Nom
khophi-lul cpffee-Acc
masi es ni]
drink Past Int
animyen if-not
[cha-lul tea-Acc masi-ess-ni]?
drink-Past-Int
‘Which of these two things did Chelswu drink: coffee or tea?’
(alt-question)
But (45b) has two other possible structural analyses available that do not involve ellipsis from an unbalanced disjunction. One is where the sub- ject in the second clause is pro,6 and the other is an analysis with VP disjunction/coordination. Each of these possibilities is represented in (49) below.
6 Like Japanese and Chinese, Korean is apro-drop language.
(49)a. [Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
masi-ess-ni]
drink Past Int
animyen if-not
[pro pro
cha-lul tea-Acc masi-ess-ni]?
drink-Past-Int
‘Which of these two things did Chelswu drink: coffee or tea?’
(alt-question) b. Chelswu-ka
Chelswu-Nom
[khophi-lul coffee-Acc
masi ess ni] drink Past Int
animyen if-not
[cha-lul tea-Acc masi-ess-ni]?
drink-Past-Int
‘Which of these two things did Chelswu drink: coffee or tea?’
(alt-question)
Let us now turn to case marking in Korean, which also provides evid- ence for the ellipsis analysis of alt-questions. In the alt-question (45b), the occurrence of the accusative case marker in bothkhophi-lul(‘coffee’) and cha-lul(‘tea’) suggests that clause disjunction is involved here rather than a simple noun phrase disjunction. As shown in theyn-question (44), a simple noun phrase disjunction allows case marking only on the head noun (i.e., the noun at the right edge of the NP). An example with a simple noun phrase conjunction also allows case marking only on the head noun, as in (50).
(50) John-i John-Nom
Mary-wa Mary-and
Sue-lul Sue-Acc
po-ass-ta.
see-Past-Decl
‘John saw Mary and Sue.’
Moreover, in (45b), each accusative-case marked noun can be followed by an adverb as illustrated in (51a), supporting our claim that it involves clause disjunction with verb deletion in the first clause. On the other hand, this is impossible in (44), illustrated in (51b), as we would expect for a simple noun phrase disjunction.
(51)a. Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
ppali quickly
animyen if-not
cha-lul tea-Acc
ppali quickly masi-ess-ni?
drink-Past-Int
‘Which of these two things did Chelswu drink quickly: coffee or tea?’ (alt-question)
b. ∗Chelswu-ka Chelswu-Nom
khophi-na coffee-or
ppali quickly
cha-lul tea-Acc
ppali quickly masi-ess-ni?
drink-Past-Int
‘Is it the case that Chelswu drank coffee or tea quickly?’ (yn- question)
Further, the case-marking facts attested in alt-questions is similar to the way case marking works in gapping in declaratives, supporting our claim that alt-questions and gapping have similar syntax. (52a) is an example of gapping in a declarative sentence, where the verb in the first conjunct has been deleted, as represented in (52b). The case marking in the first conjunct with verb deletion is the same as the second conjunct without the verb deletion: i.e., in both conjuncts, the subject NPs must bear nominative case markers and the object NPs must bear accusative case markers. A parallel example of alt-question is given in (53).
(52)a. Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
kuliko and
Sue-ka Sue-Nom
cha-lul coffee-Acc masi-ess-ta.
drink-Past-Decl
‘Chelswu drank coffee and Sue drank tea.’
b. [Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
masi ess ko]
drink-Past-Conj kuliko and
[Sue-ka Sue-Norm cha-lul
coffee-Acc
masi-ess-ta].
drink-Past-Decl
(53)a. Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
animyen if-not
Sue-ka Sue-Nom
cha-lul coffee-Acc masi-ess-ni?
drink-Past-Int
‘Which of these two is true: Chelswu drank coffee or Sue drank tea?’ (alt-question)
b. [Chelswu-ka Chelswu-Nom
khophi-lul coffee-Acc
masi-ess-ni]
cat Past Int
animyhen if-not
[Sue-ka Sue-Nom cha-lul
coffee-Acc
masi-ess-ni]?
drink-Past-Int
In sum, the data on word order and case marking in Korean show that alt-questions must disjoin full clauses (or VPs), and not just two NPs.
This provides an indirect support for ellipsis in the syntax of English alt-questions.
5.2. Focus Pattern in English
The question now is whether English has any independent evidence for ellipsis in the syntax of alt-questions. In this subsection, we argue that a suggestive piece of evidence comes from the focus pattern found in alt- questions.
A (written) question such asDid John drink coffee or tea?has, in prin- ciple, two potential readings: theyn-question reading and the alt-question reading. However, as noted in Romero (1998), focus intonation disam- biguates the two readings: the yn-reading presents neutral intonation on the disjunctive phrase, as in (54a), whereas the alt-reading is in general achieved by placing focus stress on each disjunct, as in (54b).
(54)a. Did John drink coffee or tea? (neutral intonation, yn-reading only)
b. Did John drink COFfee or TEA? (focus in capitals, alt-reading only)
In a question with gapping in the second clause, as in (55), the remnant and the correlate are focused and the question only has the alt-reading.
(55) Did JOHN drink COFfee or MARY TEA?
Also, a yn-reading in a question with overt or not correlates with the presence of stress on the verb and onnot, as in (56).
(56) Did John DRINK or NOT?
What is the function of this double focus? A common use of focal stress is to signal that the sentence with the focus is semantically parallel to some other nearby sentence, contrasting only in the focused part.8For example, in (57), the two disjunctive clauses are parallel to each other, differing only in the content of the NP objects. Similarly, in (58), the two adjacent clauses naturally bear stress on the non-common elements, namely on the contrastive NP subjects and on the contrastive temporal adjuncts. Finally,
8 For a formalization of the felicity conditions of contrastive focus, see Rooth (1985, 1992).
the same holds for (59), where the polarities and the embedded predicates differ from each other and bear focal stress:
(57) Either [IP Sita ate BEEF for dinner] or [IP she ate PORK for dinner].
(58) [IPPAT visited Sue for CHRISTmas], and [IPJOHN visited Sue for NEW YEAR].
(59) [IP Noa does NOT play tennis when the ground is WET]. [IP
She PLAYS it when the ground is DRY].
In cases where disjoined sentences involve ellipsis in the second clause, the remnant material must necessarily bear some focal stress. This necessary focus on the remnant is typically mirrored with focus stress on the correl- ate, since the remnant and its correlate are the parts that are contrasting with each other.
(60)a. MaTILda WENT to the theatre, but SUsan did NOT go to the theater.
b. JOHN drank COFfe and MARY drank TEA.
We can now see that a similar analysis can be applied to whether/Q . . . or constructions, naturally explaining the double focus, if they also involve disjunction of clauses, and in some cases ellipsis in the second clause. For instance, in (61b), each of the constituents in IP1 contrasts with the corresponding constituent in IP2, hence focus stress is placed on all of them.9
(61)a. Did MARY SING or BILL DANce?
b. Qi Didti [IP1MARY SING] or [IP2BILL DANce]?
Further, ifwhether/Q . . . orconstructions are clausal disjunctions, then the examples in (54b) and (56), as well as the example in (55), all involve ellipsis in the second clause. The resulting structures are illustrated in (62).
For these examples, the focus stress falls on the remnant and its correlate,
9 We thank a reviewer for raising the issue of focus pattern in alt-questions that do not involve ellipsis, and for pointing us to relevant examples.
as is typical in elliptical constructions like VP-ellipsis (as in (60a)) and gapping in declaratives (as in (60b)).
(62)a. Qi Didti [I P John drink COFfee] or [I P he drink TEA]?
b. Qi Didti [I P John DRINK] or [I P he NOT drink]?
c. Qi Didti [I P JOHN drink COFfee] or [I P MARY drink TEA]?
Under the movement-only analysis, there is no straightforward way to account for the difference in the focus pattern between the yn- and the alt-reading. For instance, the example in (14) (repeated here as (63)), has two possible derivations under the movement-only analysis, depending on whether it has theyn- or the alt-reading.
(63) Did John drink coffee or tea?
a. yn-question:
Opi (ti or not) [did John drink [Opj coffee or tea]]
b. alt-question:
Opi [did John drink [ti COFfee or TEA]]
In both derivations, coffee or teais associated with an operator. The dif- ference is that in the alt-reading, the operator associated with it moves to COMP, while in the yn-reading, it does not move. It is not clear how we can link theyn-reading configuration to no accent and the alt-reading configuration to the double accent focus pattern.
The same point can be raised foreither . . . orconstructions. According to the native speakers that we consulted, while clausaleither . . . orin (64b, c) necessarily triggers double focus on the disjuncts, either NP or NP in (64a) doesn’t necessarily do so.10
(64)a. John drank either coffee or tea.
b. Either John drank COFfee or TEA.
c. John either drank COFfee or TEA.
To further verify this intuition, consider the context in (65). Although the judgments are subtle, all the native speakers we consulted agreed that
10 For some speakers, it was possible to put double focus on the disjuncts in (64a) in some contexts, although not as strong as in (64b, c). The crucial point for us is that while sentences witheitherplaced at VP or at IP level necessarily carry double focus on the disjuncts, those witheitherplaced at the NP level do not necessarily do so.
while (65a) is a perfect continuation, (65b) is funny. Note that the scenario is set up in such a way that the constituents being contrasted are drank and ate, and not coffee and tea. This means that while drank and ate must be stressed, coffee and tea should not be. This intonational pattern is compatible with (65a) because the NP disjunctioncoffee or teain (65a) does not necessarily have double focus. However, in (65b), coffee or tea necessarily has double focus – since it involves clausal disjunction plus ellipsis –, which is in conflict with the intonational pattern called for by the context.
(65) CONTEXT: John is discovered dead in his office 15 minutes after the lunch he attended is over. He’s been poisoned. The police suspects that the poison was in the food and they are trying to find out what he ate. Then A says:
a. A: John drank either coffee or tea, but I’ve no idea what he ate.
b. #A: John either drank coffee or tea, but I’ve no idea what he ate.
In sum, nothing in the movement-only analysis per se predicts an intonational difference between disjunction that moves to a clausal site (movement analysis for (63b) and (64b, c)) and disjunction that remains at the NP level (examples (63a) and (64a)). Under the ellipsis analysis, instead, the necessary double focus on the disjuncts in (63b) and (64b, c) can be explained as the focus that signals the necessary contrast between a remnant and its correlate in ellipsis. Focal stress in (63a) and (64a) is optional and is probably determined by discourse factors, depending on whether or not the disjuncts are meant to be used contrastively in that context.
5.3. Summary of Section 5
We started this section by pointing out that, once whether/Q . . . oris re- cognized as thewh-version ofeither . . . or, and given that coordination is crosscategorial and gapping is available in the grammar, an alt-question like (66) can in principle have the derivations in (66a) and in (66b). The question was: does one of the two derivations win over the other and, if so, which one?
(66) Did John drink coffee or tea?
a. Qi Did John drinkti [coffee or tea]
b. Qi Didti [John drink coffee or John drink tea]
We have seen that the string syntax of Hindi and Korean alt-questions can only be explained if the derivation of the alt-reading involves disjunc- tion of two clausal nodes, as in (66b), and not coordination of two NPs, as in (66a). That is, the crosslinguistic data show that the ellipsis/movement derivation in (66b) must win over the pure movement derivation in (66a).
Data on the phonological focus pattern ofwhether/Q . . . orconstructions in English also suggest that these questions must also be derived via ellipsis plus movement. In view of these data, we conclude thatwhether/Q . . . or constructions are the wh-version of clausal either . . . or constructions, hence involving both ellipsis and movement.
A question still remains as to why in alt-questions the derivation in- volving clausal disjunction with ellipsis wins over the pure movement derivation. We do not have a full answer to this question. However, note that for an alt-question like (66) the movement in the combined move- ment/ellipsis derivation in (66b) is shorter than the one in the movement- only derivation in (66a). An economy principle preferring Shortest Move may point us towards an answer: it may be that, for a given reading, shorter movement from disjunction of larger constituents (with consequent ellipsis) is preferred over longer movement from disjunction of smaller constituents.
6. COMBINATION OF ELLIPSIS AND MOVEMENT INALTERNATIVE
QUESTIONS
Incorporating insights from both Larson and Schwarz, we have so far argued that the syntax of whether/Q . . . or constructions like (1) and (3) involves disjunction of clauses with consequent ellipsis in the second clause, as well as movement of whether/Qassociated with a disjunctive clause to the matrix [Spec, CP]. In this section, we will first show how our combined movement/ellipsis analysis handles the apparent asymmetries betweeneither . . . orandwhether/Q . . . ordiscussed in section 3. We will then briefly consider possible analyses of yn-questions without an overt associated disjunction.
6.1. Asymmetries between either . . . or and whether/Q . . . or
Recall from section 3 that while either . . . or constructions do not al- low dangling remnants and ellipsis across finite clauses, whether/Q . . . or constructions seem to do so. The examples illustrating these apparent
asymmetries are given in (67)–(69), with the corresponding derivations under a pure ellipsis account.
(67)a.??Either this pissed Bill or Sue off.
Either [IPthis pissed Bill] or [IPthis pissed Sue off]
b. Did this piss Bill or Sue off?
Q [C Did this piss Bill] or [C Did this piss Sue off]
(68)a.??Either John said that Bill retired or resigned.
Either [IPJohn said that Bill retired] or [IPJohn said that Bill resigned]
b. Did John say that Bill retired or resigned?
Q [C Did John say that Bill retired] or [C Did John say that Bill resigned]
(69)a.??Either John claimed that Bill drank coffee or tea.
Either [IPJohn claimed that Bill drank coffee] or [IPBill claimed that John drank tea]
b. Did John claim that Bill drank coffee or tea?
Q [C Did John claim that Bill drank coffee] or [IPDid Bill claim that John drank tea]
The starting point of our analysis is that the degradedeither . . . orcon- structions with a dangling remnant and those with ellipsis spreading across matrix and finite embedded clauses become well-formed ifeitheris lower in the clause.
(70)a. This either pissed Bill or Sue off.
b. This either [VPpissed Billej] or [VPpissed Sueej] offj
(71)a. John said that Bill either retired or resigned.
b. John said that Bill either [VPretired] or [VPresigned]
(72)a. John claimed that either Bill drank coffee or tea.
b. John said that either [IPBill drank coffee] or [IPBill drank tea]
According to Schwarz, in (70a),eitheris adjoined to VP marking the left periphery of the first disjunct, and the particleoffhas undergone right-node raising, as represented in (70b). Given this analysis, the only elided mater- ial is the verb pissedin the second disjunct, and so there is no dangling remnant. Schwarz further notes that in (67a) (repeated here as (73a)) the option of right-node raising the particle, as in (73b), is difficult, if not completely unavailable, because the particle would have to right-node raise above IP. But the right-node raising option is available for (70a) because the particle is required to raise only above VP.
(73)a.??Either this pissed Bill or Sue off.
b.??Either [IPthis pissed Billej] or [IPthis pissed Sueej] offj. In (71a), either is adjoined to the embedded disjunctive VP, and so there is no ellipsis involved, as represented in (71b). In (72a), either is adjoined to the embedded disjunctive IP, with ellipsis in the second IP.
Both sentences are well-formed with the reading in which the scope ofor is restricted to the embedded clause.
The difference between whether/Q and either is that whether/Q is a wh-phrase, whileeitheris not, and sowhether/Qcan undergo movement, leaving a trace, whileeithercannot. In other words, the trace ofwhether/Q corresponds to the surface position ofeither. This means that whileeither marks the left edge of the disjunction in either . . . or constructions, the trace of whether/Qmarks the left edge of disjunction inwhether/Q . . . or constructions. This in turn means that the left edge of ellipsis can be the originating position ofwhether/Q, and not its surface position, and that as long as there is a grammatical source sentence witheither, the correspond- ingwhether/Qsentence should be well-formed. Further, the movement of whether/Qto matrix [Spec, CP] will expand the scope ofor, allowing it to have the widest scope. This then is whywhether/Q . . . orconstructions ap- pear to allow dangling remnants and ellipsis across matrix and embedded finite clauses. The LFs we propose for (67b)–(69b) are given in (74a)–
(74c) respectively. For us, (67b) does not involve a dangling remnant, and (68b)–(69b) do not involve ellipsis across matrix and finite embedded clauses.
(74)a. Qi Did thisti [VPpiss Billej] or [VPpiss Sueej] offj
b. Qi Did John say that Billti [VPretired] or [VPresigned]
c. Qi Did John claim thatti [IPBill drank coffee] or [IPBill drank tea]
In our analysis, the apparent ellipsis from unbalanced disjunction actually involves disjunction of VPs with ellipsis in the second VP and a right-node raised particle. The apparent ellipsis across matrix and embedded finite clauses also involves disjunction of VPs or IPs. Further, in this case, the movement of Qto matrix [Spec, CP] allowsor to have the widest scope, giving rise to the alt-reading.
Two questions concerning the contrast between (70a) and (73a) (re- peated below as (75a) and (75b)) need to be addressed at this point.11
(75)a.??Either this pissed Bill or Sue off.
b. This either pissed Bill or Sue off.
First, is there an alternative non-ellipsis account of the contrast in (75) that makes the same prediction as Schwarz’s gapping and right-node raising analysis? A reviewer suggests that an alternative analysis may be possible if the strategy proposed in Kayne (1998) for the syntactic derivation of negative sentences like (76) is adopted.
(76) I forced us to turn no one down.
In (76),no onecannot have wide scope. According to Kayne, the derivation that obtains wide scope reading ofno onewould involve movement ofno oneto [Spec, NegP] in the matrix clause, followed by particle raising and remnant VP movement, as in (77). But the problem is that the particle would have to raise out of the embedded clause and undergo a long- distance movement to the matrix clause. For Kayne, this is illegal, and hence (76) cannot be associated with a wide scope reading ofno one.
(77) Source: I forced us [to turn no one down].
a. Neg preposing:
I no onei forced us [to turnti down]
b. Particle preposing:∗
I no onei downkforced us [to turntitk] c. Remnant VP movement:
I [VPforced us to turntitk]j no onei downktj
In a nutshell, if we were to apply Kaynean strategy to the examples in (75), the derivations would involve particle raising and either NP or
11 We thank the reviewers for raising these important questions.
NPraising, followed by a remnant VP/IP movement. The derivations for (75a) and (75b) will roughly proceed as in (78) and (79) respectively. The asymmetry in the grammaticality between (75a) and (75b) can then be attributed to the difference in the height of the particle raising: in (75a), the particle raises above IP as in (78a), while in (75b), it raises above VP as in (79a).
(78) Source: [IPthis [VPpissed either [Bill or Sue] off]]
a. Particle preposing: ??
offi [IPthis pissed either Bill or Sueti] b. Disjunctive NP movement:
[either Bill or Sue]j offi [IPthis pissedtjti] c. Remnant IP movement:
either [IPthis pissedtjti]k[Bill or Sue]j offitk
(79) Source: this [VPpissed either [Bill or Sue] off]
a. Particle preposing:
this offi [VPpissed either Bill or Sueti] b. Disjunctive NP movement:
this [either Bill or Sue]j offi [VPpissedtjti] c. Remnant VP movement:
this either [VPpissedtjti]k [Bill or Sue]j offitk
Note however that in order to derive the correct string order, we are forced to insert the remnant VP/IP in between either and NP or NP. It is not clear to us what kind of syntactic operation would motivate this.
Further, the Kaynean strategy cannot be extended to Hindi and Korean.
This is because the disjuncts in Hindi and Korean alt-questions do not form contiguous constituents. Admittedly, more thorough research needs to be done to gain full understanding of what all this means. We leave this open for future research.
The second question is concerned with the unmovability of clausal either. It is crucial for Schwarz’s analysis that clausaleitherdoes not move, for otherwise (80a) could have the parse in (80b) and, since the particle only raises to VP, (80a) would have the same grammaticality status as (81a) (and as (82a)), contrary to fact.12
12 A reviewer notes that for some speakers, the example in (80a) is somewhat ok, with emphasis, in certain discourse contexts, and suggests that this example cast doubt on the
